arXiv subjects

Tao Wang

Publications and source records attributed to Tao Wang.

At least 775 records · Page 43Linked to original sources

arXiv2019

Cluster Pairwise Error Probability and Construction of Parity-Check-Concatenated Polar Codes

A successive cancellation list (SCL) decoder with limited list size for polar codes can not be analyzed as a successive cancellation (SC) decoder, nor as a maximum likelihood (ML) decoder, due to the complicated decoding errors caused by path elimination. To address this issue, an analytical tool, named as cluster pairwise error probability (CPEP), is proposed in this paper to measure the competitiveness of the correct path against the error paths in an SCL decoder. It is shown that the sum of CPEPs over error paths could be used as an indicator of the probability of correct path being eliminated from the decoder list. Then, we use CPEP to explain the error performance gain of parity-check-concatenated (PCC) polar code, and apply CPEP as the optimization criterion in the construction of PCC polar codes, aiming to reduce the elimination probability of the correct path in an SCL decoder with limited list size. Simulation results show that the constructed CRC-PCC polar codes outperform their counterparts of CRC-concatenated polar codes over various codeword lengths, code rates and puncturing patterns.

cs.IT↗

arXiv2019

Few-shot Adaptive Faster R-CNN

To mitigate the detection performance drop caused by domain shift, we aim to develop a novel few-shot adaptation approach that requires only a few target domain images with limited bounding box annotations. To this end, we first observe several significant challenges. First, the target domain data is highly insufficient, making most existing domain adaptation methods ineffective. Second, object detection involves simultaneous localization and classification, further complicating the model adaptation process. Third, the model suffers from over-adaptation (similar to overfitting when training with a few data example) and instability risk that may lead to degraded detection performance in the target domain. To address these challenges, we first introduce a pairing mechanism over source and target features to alleviate the issue of insufficient target domain samples. We then propose a bi-level module to adapt the source trained detector to the target domain: 1) the split pooling based image level adaptation module uniformly extracts and aligns paired local patch features over locations, with different scale and aspect ratio; 2) the instance level adaptation module semantically aligns paired object features while avoids inter-class confusion. Meanwhile, a source model feature regularization (SMFR) is applied to stabilize the adaptation process of the two modules. Combining these contributions gives a novel few-shot adaptive Faster-RCNN framework, termed FAFRCNN, which effectively adapts to target domain with a few labeled samples. Experiments with multiple datasets show that our model achieves new state-of-the-art performance under both the interested few-shot domain adaptation(FDA) and unsupervised domain adaptation(UDA) setting.

cs.CV↗

arXiv2019

A Unified Analytical Framework for Trustable Machine Learning and Automation Running with Blockchain

Traditional machine learning algorithms use data from databases that are mutable, and therefore the data cannot be fully trusted. Also, the machine learning process is difficult to automate. This paper proposes building a trustable machine learning system by using blockchain technology, which can store data in a permanent and immutable way. In addition, smart contracts are used to automate the machine learning process. This paper makes three contributions. First, it establishes a link between machine learning technology and blockchain technology. Previously, machine learning and blockchain have been considered two independent technologies without an obvious link. Second, it proposes a unified analytical framework for trustable machine learning by using blockchain technology. This unified framework solves both the trustability and automation issues in machine learning. Third, it enables a computer to translate core machine learning implementation from a single thread on a single machine to multiple threads on multiple machines running with blockchain by using a unified approach. The paper uses association rule mining as an example to demonstrate how trustable machine learning can be implemented with blockchain, and it shows how this approach can be used to analyze opioid prescriptions to help combat the opioid crisis.

cs.LG↗

arXiv2019

NetKernel: Making Network Stack Part of the Virtualized Infrastructure

This paper presents a system called NetKernel that decouples the network stack from the guest virtual machine and offers it as an independent module. NetKernel represents a new paradigm where network stack can be managed as part of the virtualized infrastructure. It provides important efficiency benefits: By gaining control and visibility of the network stack, operator can perform network management more directly and flexibly, such as multiplexing VMs running different applications to the same network stack module to save CPU cores, and enforcing fair bandwidth sharing with distributed congestion control. Users also benefit from the simplified stack deployment and better performance. For example mTCP can be deployed without API change to support nginx natively, and shared memory networking can be readily enabled to improve performance of colocated VMs. Testbed evaluation using 100G NICs shows that NetKernel preserves the performance and scalability of both kernel and userspace network stacks, and provides the same isolation as the current architecture.

cs.NI↗

arXiv2019

ALMA twenty-six arcmin$^2$ survey of GOODS-S at one-millimeter (ASAGAO): Near-infrared-dark faint ALMA sources

We report detections of two 1.2 mm continuum sources ($S_\mathrm{1.2mm}$ ~ 0.6 mJy) without any counterparts in the deep $H$- and/or $K$-band image (i.e., $K$-band magnitude $\gtrsim$ 26 mag). These near-infrared-dark faint millimeter sources are uncovered by ASAGAO, a deep and wide-field ($\simeq$ 26 arcmin$^2$) Atacama Large Millimeter/submillimeter Array (ALMA) 1.2 mm survey. One has a red IRAC (3.6 and 4.5 $μ$m) counterpart, and the other has been independently detected at 850 and 870 $μ$m using SCUBA2 and ALMA Band 7, respectively. Their optical to radio spectral energy distributions indicate that they can lie at $z \gtrsim$ 3-5 and can be in the early phase of massive galaxy formation. Their contribution to the cosmic star formation rate density is estimated to be ~ 1 $\times$ 10$^{-3}$ $M_\odot$ yr$^{-1}$ Mpc$^{-3}$ if they lie somewhere in the redshift range of $z$ ~ 3-5. This value can be consistent with, or greater than that of bright submillimeter galaxies ($S_\mathrm{870μm}>$ 4.2 mJy) at $z$ ~ 3-5. We also uncover 3 more candidates near-infrared-dark faint ALMA sources without any counterparts ($S_\mathrm{1.2mm}$ ~ 0.45-0.86 mJy). These results show that an unbiased ALMA survey can reveal the dust-obscured star formation activities, which were missed in previous deep optical/near-infrared surveys.

astro-ph.GA↗

arXiv2019

Real-Time Fine-Grained Air Quality Sensing Networks in Smart City: Design, Implementation and Optimization

Driven by the increasingly serious air pollution problem, the monitoring of air quality has gained much attention in both theoretical studies and practical implementations. In this paper, we present the architecture, implementation and optimization of our own air quality sensing system, which provides real-time and fine-grained air quality map of the monitored area. As the major component, the optimization problem of our system is studied in detail. Our objective is to minimize the average joint error of the established real-time air quality map, which involves data inference for the unmeasured data values. A deep Q-learning solution has been proposed for the power control problem to reasonably plan the sensing tasks of the power-limited sensing devices online. A genetic algorithm has been designed for the location selection problem to efficiently find the suitable locations to deploy limited number of sensing devices. The performance of the proposed solutions are evaluated by simulations, showing a significant performance gain when adopting both strategies.

cs.OH↗

arXiv2019

Anomalous Spin-Orbit Torques in Magnetic Single-Layer Films

Spin-orbit interaction (SOI) couples charge and spin transport, enabling electrical control of magnetization. A quintessential example of SOI-induced transport is the anomalous Hall effect (AHE), first observed in 1880, in which an electric current perpendicular to the magnetization in a magnetic film generates charge accumulation on the surfaces. Here we report the observation of a counterpart of the AHE that we term the anomalous spin-orbit torque (ASOT), wherein an electric current parallel to the magnetization generates opposite spin-orbit torques on the surfaces of the magnetic film. We interpret the ASOT as due to a spin-Hall-like current generated with an efficiency of 0.053+/-0.003 in Ni80Fe20, comparable to the spin Hall angle of Pt. Similar effects are also observed in other common ferromagnetic metals, including Co, Ni, and Fe. First principles calculations corroborate the order of magnitude of the measured values. This work suggests that a strong spin current with spin polarization transverse to magnetization can exist in a ferromagnet, despite spin dephasing. It challenges the current understanding of spin-orbit torque in magnetic/nonmagnetic bilayers, in which the charge-spin conversion in the magnetic layer has been largely neglected.

cond-mat.mes-hall↗

arXiv2019

Bernstein Polynomial Model for Nonparametric Multivariate Density

In this paper, we study the Bernstein polynomial model for estimating the multivariate distribution functions and densities with bounded support. As a mixture model of multivariate beta distributions, the maximum (approximate) likelihood estimate can be obtained using EM algorithm. A change-point method of choosing optimal degrees of the proposed Bernstein polynomial model is presented. Under some conditions the optimal rate of convergence in the mean $χ^2$-divergence of new density estimator is shown to be nearly parametric. The method is illustrated by an application to a real data set. Finite sample performance of the proposed method is also investigated by simulation study and is shown to be much better than the kernel density estimate but close to the parametric ones.

stat.ME↗

arXiv2019

GM-PLL: Graph Matching based Partial Label Learning

Partial Label Learning (PLL) aims to learn from the data where each training example is associated with a set of candidate labels, among which only one is correct. The key to deal with such problem is to disambiguate the candidate label sets and obtain the correct assignments between instances and their candidate labels. In this paper, we interpret such assignments as instance-to-label matchings, and reformulate the task of PLL as a matching selection problem. To model such problem, we propose a novel Graph Matching based Partial Label Learning (GM-PLL) framework, where Graph Matching (GM) scheme is incorporated owing to its excellent capability of exploiting the instance and label relationship. Meanwhile, since conventional one-to-one GM algorithm does not satisfy the constraint of PLL problem that multiple instances may correspond to the same label, we extend a traditional one-to-one probabilistic matching algorithm to the many-to-one constraint, and make the proposed framework accommodate to the PLL problem. Moreover, we also propose a relaxed matching prediction model, which can improve the prediction accuracy via GM strategy. Extensive experiments on both artificial and real-world data sets demonstrate that the proposed method can achieve superior or comparable performance against the state-of-the-art methods.

cs.LG↗

arXiv2018

Long-distance continuous-variable measurement-device-independent quantum key distribution with discrete modulation

We propose a long-distance continuous-variable measurement-device-independent quantum key distribution (CV-MDI-QKD) protocol with discrete modulation. This kind of discrete-modulated schemes have good compatibility with efficient error correction code, which lead to higher reconciliation efficiency even at low signal-to-noise ratio (SNR). Security analysis shows that the proposed protocol is secure against arbitrary collective attacks in the asymptotic limit with proper use of decoy states. And with the using of discrete modulation, the proposed CV-MDI-QKD protocol has simpler implementation and outperform previous protocols in terms of achievable maximal transmission distance, which precisely solve the bottleneck of the original Gaussian-modulated CV-MDI-QKD protocol.

quant-ph↗

arXiv2018

Image Classification at Supercomputer Scale

Deep learning is extremely computationally intensive, and hardware vendors have responded by building faster accelerators in large clusters. Training deep learning models at petaFLOPS scale requires overcoming both algorithmic and systems software challenges. In this paper, we discuss three systems-related optimizations: (1) distributed batch normalization to control per-replica batch sizes, (2) input pipeline optimizations to sustain model throughput, and (3) 2-D torus all-reduce to speed up gradient summation. We combine these optimizations to train ResNet-50 on ImageNet to 76.3% accuracy in 2.2 minutes on a 1024-chip TPU v3 Pod with a training throughput of over 1.05 million images/second and no accuracy drop.

cs.LG↗

arXiv2018

Total coloring of 1-toroidal graphs of maximum degree at least 11 and no adjacent triangles

A {\em total coloring} of a graph $G$ is an assignment of colors to the vertices and the edges of $G$ such that every pair of adjacent/incident elements receive distinct colors. The {\em total chromatic number} of a graph $G$, denoted by $\chiup''(G)$, is the minimum number of colors in a total coloring of $G$. The well-known Total Coloring Conjecture (TCC) says that every graph with maximum degree $Δ$ admits a total coloring with at most $Δ+ 2$ colors. A graph is {\em $1$-toroidal} if it can be drawn in torus such that every edge crosses at most one other edge. In this paper, we investigate the total coloring of $1$-toroidal graphs, and prove that the TCC holds for the $1$-toroidal graphs with maximum degree at least~$11$ and some restrictions on the triangles. Consequently, if $G$ is a $1$-toroidal graph with maximum degree $Δ$ at least~$11$ and without adjacent triangles, then $G$ admits a total coloring with at most $Δ+ 2$ colors.

math.CO↗

arXiv2018

Minor stars in plane graphs with minimum degree five

The weight of a subgraph $H$ in $G$ is the sum of the degrees in $G$ of vertices of $H$. The {\em height} of a subgraph $H$ in $G$ is the maximum degree of vertices of $H$ in $G$. A star in a given graph is minor if its center has degree at most five in the given graph. Lebesgue (1940) gave an approximate description of minor $5$-stars in the class of normal plane maps with minimum degree five. In this paper, we give two descriptions of minor $5$-stars in plane graphs with minimum degree five. By these descriptions, we can extend several results and give some new results on the weight and height for some special plane graphs with minimum degree five.

math.CO↗

arXiv2018

Exploration of Micro-VCSEL Ultra-low Biasing Scheme for Pulse Generation

The generation of optical pulses at ultra low bias level, thus low energy cost, is explored in a commercial microcavity semiconductor laser in view of testing the principle of energy efficient information encoding in potential integrated schemes. Sequences of regular, highly nonlinear pulses with acceptable amplitude stability are obtained from a commercial device as potential sources of bits where the information is added by post-treatment (pulse removal). A discussion on the energy expenditure per bit is offered, together with the optimal frequency for pulse generation, which is found to lie slightly below the above-threshold value declared by the manufacturer.

physics.optics↗

arXiv2018

Precision metrology of weak measurement with thermal state pointer

Quantum metrology is being gradually studied for weak measurement systems. For weak measurement systems with thermal state pointer, we find that in the displacement space corresponding to imaginary weak values, the maximal QFI after successful postselection can attain the level of thermal fluctuations, without surpassing total QFI, and that QFI which increases with increasing temperature can constantly improve the measurement precision. These results are much better than that of weak measurement with pure state (i.e., Gaussian state) pointer. On the other hand, in Kerr nonlinear interaction systems with weak measurement, and by using thermal state pointer, we obtain in the phase space successful postselection and postselected measurements both achieve the Heisenberg limit of quantum metrology, and show weak measurement with thermal states only obtain classical Fisher information (CFI) which increases with increasing temperature and achieves classical enhanced scaling of N^2. Moreover, weak measurement with thermal states has an advantage over that with coherent states or mixed states of the light because generating these states with more large uncertainty are limited under the current technology, but thermal states with more large uncertainly are very easy to achieve with increasing temperature in nature, regardless of thermal states of the light or the matter.

quant-ph↗

arXiv2018

Two-Dimensional Vortex Sheets for the Nonisentropic Euler Equations: Nonlinear Stability

We show the short-time existence and nonlinear stability of vortex sheets for the nonisentropic compressible Euler equations in two spatial dimensions, based on the weakly linear stability result of Morando--Trebeschi (2008) [20]. The missing normal derivatives are compensated through the equations of the linearized vorticity and entropy when deriving higher-order energy estimates. The proof of the resolution for this nonlinear problem follows from certain \emph{a priori} tame estimates on the effective linear problem {in the usual Sobolev spaces} and a suitable Nash--Moser iteration scheme.

math.AP↗

arXiv2018

Revealing environmental dependence of molecular gas content in a distant X-ray cluster at z=2.51

We present a census of the molecular gas properties of galaxies in the most distant known X-ray cluster, CLJ1001, at z=2.51, using deep observations of CO(1-0) with JVLA. In total 14 cluster members with $M_{*} > 10^{10.5} M_{\odot}$ are detected, including all the massive star-forming members within the virial radius, providing the largest galaxy sample in a single cluster at $z > 2$ with CO(1-0) measurements. We find a large variety in the gas content of these cluster galaxies, which is correlated with their relative positions (or accretion states), with those closer to the cluster core being increasingly gas-poor. Moreover, despite their low gas content, the galaxies in the cluster center exhibit an elevated star formation efficiency (SFE=SFR/$M_{\rm gas}$) compared to field galaxies, suggesting that the suppression on the SFR is likely delayed compared to that on the gas content. Their gas depletion time is around $t_{\rm dep} \sim 400$ Myrs, comparable to the cluster dynamical time. This implies that they will likely consume all their gas within a single orbit around the cluster center, and form a passive cluster core by $z\sim2$. This result is one of the first direct pieces of evidence for the influence of environment on the gas reservoirs and SFE of $z > 2$ cluster galaxies, thereby providing new insights into the rapid formation and quenching of the most massive galaxies in the early universe.

astro-ph.GA↗

arXiv2018

Atmospheric effects on continuous-variable quantum key distribution

Compared to fiber continuous-variable quantum key distribution (CVQKD), atmospheric link offers the possibility of a broader geographical coverage and more flexible transmission.However, there are many negative features of the atmospheric channel that will reduce the achievable secret key rate, such as beam extinction and a variety of turbulence effects. Here we show how these factors affect performance of CVQKD, by considering our newly derived key rate formulas for fading channels,which involves detection imperfections, thus form a transmission model for CVQKD. This model can help evaluate the feasibility of experiment scheme in practical applications.We found that performance deterioration of horizontal link within the boundary layer is primarily caused by transmittance fluctuations (including beam wandering, broadening, deformation, and scintillation),while transmittance change due to pulse broadening under weak turbulence is negligible. Besides,we also found that communication interruptions can also cause a perceptible key rate reduction when the transmission distance is longer, while phase excess noise due to arrival time fluctuations requires new compensation techniques to reduce it to a negligible level. Furthermore, it is found that performing homodyne detection enables longer transmission distances, whereas heterodyne allows higher achievable key rate over short distances.

quant-ph↗